Stable relative to what
A stablecoin is a token engineered to hold steady value against some reference. The reference is usually the US dollar but not always. The chart below shows the relative scale of all six categories. Dollar-pegged coins dominate, but the taxonomy runs wider than the common case this dashboard tracks.
Sizing comparison, not non-overlapping global market shares. These categories use different historical dates and definitions, and some can overlap. Multi-value estimates use their midpoint. RWA value is a dated secondary report excluding stablecoins, not current primary fund AUM. Any percentages in the chart describe its own bars, not the stablecoin market as a whole.
The six categories
Token comparison
Sortable by market cap; filter by category with the chips. Figures as of mid-2026.
| Token | Category | Peg | Issuer | Market cap | Chain | As of |
|---|
From crypto plumbing to financial infrastructure
One dated measurement puts total stablecoin market cap at $308.0B as of 13 August 2026, up 14.3% year over year, about 99.5% dollar-denominated and 4.5% below its May 2026 peak. That is one aggregator's number for one date. No second independent measurement of the same date and universe is verified in this report's research record, so no cross-source range is quoted and the absence is stated rather than filled. The projections below are scenarios from different forecasters, not a consensus or a guarantee of multi-trillion scale.
Total stablecoin market cap, USD billions. The 2017-2025 points are year-end observations; the 2026 point is a dated observation, not a year-end total. Markers: UST collapse (May 2022), USDC/SVB (March 2023), GENIUS Act signed (July 2025). Source: DefiLlama stablecoins dashboard.
Forward projections to 2030
Named projections to 2030, not a consensus forecast. Toggle forecasters with the chips below the chart. Single-point forecasts are shown with small visual padding so the bar is visible; see tooltip for the exact stated figure.
A quiet buyer of government debt
Stablecoin issuers hold well over $150B in US Treasuries, mostly short-dated T-bills. Tether's reported ~$141B (Q1 2026 attestation, exposure as of 2026-03-31) exceeds the holdings of Germany ($91.3B) and Norway ($104.4B) per the US Treasury TIC Jan 2023 table; combined issuers' ~$182B exceeds Norway. The comparison is duration-mismatched: issuers hold short-dated bills, while TIC ranks total foreign holdings. And the figure above is gross exposure, not incremental demand: an issuer buying bills only adds to Treasury demand if nobody else sold them to fund the purchase. The section below separates the two.
Largest holders of US Treasury securities, USD billions. Sovereign top 3 per US Treasury TIC Jun 2026; remaining sovereign rows per TIC Jan 2023 pending refresh; stablecoin issuers per their own reserve attestations (Tether Q1 2026, combined 4 issuers ~$182B). The comparison is duration-mismatched - issuers hold short-dated T-bills, TIC ranks total (long+short) holdings.
The GENIUS Act was enacted July 18, 2025 (the House passed it July 17, so "July 17-18" appears in coverage). Section 4 requires payment-stablecoin reserves to be at least 1:1, and the permitted assets are a list of eight classes, not Treasury bills alone: coins and currency; insured deposits; Treasury bills, notes or bonds of 93 days or less (the statute says 93, not 90); overnight repo backed by such securities; overnight reverse repo against Treasuries; shares of a registered Government money market fund invested solely in the preceding classes; other similarly liquid Federal Government-issued assets approved by the regulator; and tokenized forms of those classes. Section 4 also requires monthly published reserve composition including average tenor and custody location per category, and bars pledging or rehypothecating reserves outside its exceptions. The full list is reproduced in Section 8.
Commencement is not settled. Section 20 sets general commencement at the earlier of 18 months after enactment or 120 days after the primary federal payment stablecoin regulators issue any final implementing regulations. January 18, 2027 is therefore an outer limit, not a fixed date, and whether a final rule has already triggered the earlier date is UNRESOLVED in this report's research record. Section 3(b)(1) separately places the offer-and-sale prohibition three years after enactment. This report therefore does not describe any GENIUS duty as locked, operative or currently in force. Re-verify commencement before relying on these statutory duties.
See the full regulatory comparison in Section 8.
The eight eligible reserve classes
Stated as the statute states it. Reading "1:1 reserves" as "1:1 Treasury bills" is the specific error this list is here to prevent.
Gross holdings are not net new demand
The clearest available example of why, and the place where the cited source disagrees with itself.
T-bill maturity bands: why issuers stay short
Reserve composition by maturity, from issuer attestations. Short-term bills (93 days or less, the statutory cap) dominate because redemptions can arrive any day; long-dated holdings would force a fire sale.
Sources: Tether Q1 2026 attestation; Circle Q1 2026 reserve report. Yadav and Malone (JIEL 2025) flag the interdependence: mass redemptions of stablecoins could strain short-term Treasury liquidity.
Does this take money out of banks
When a household moves a deposit into a stablecoin, the bank loses cheap funding that funds mortgages and business loans. The research is genuinely contested, not settled. The diagram below shows the mechanism; the callouts below it give both sides, cited and neutral.
A $1 deposit moved to a stablecoin issuer does not disappear - it becomes a Treasury holding instead of bank lending capital. Distribution matters, not just aggregate liquidity.
Banks exposed to stablecoin flows lend less
NY Fed, Staff Report 1185 (Feb 2026)
First direct evidence of liquidity-driven disintermediation. Partner banks run "narrow" to absorb the volatility of stablecoin flows; their loan-to-asset ratios contract versus peers. The ECB (Schnabel, Jun 2026) draws the parallel to 1970s money-market funds draining retail deposits.
Minimal lending impact modeled
White House CEA (Apr 2026)
Assumes a small baseline market. Criticized by Americans for Financial Reform ("a model built to mislead," May 2026) and the Consumer Bankers Association (Apr 2026) for treating recirculated funds as equivalent to retail deposits when they return as concentrated, wholesale, rate-sensitive liabilities that do not support relationship lending.
Both sides are politically situated. The honest answer right now: early evidence, no consensus number.
The yield debate
Should stablecoins pay interest? The GENIUS Act says no for payment stablecoins; the research on yield-bearing variants is contested.
Sources: CRS IF13173/IF13174; State Street (Apr 2026); BPI (2026, citing Cong, Chiu et al.); Federal Reserve FEDS Notes (Dec 2025).
Cheaper, faster money across borders
Cross-border cost is not one number. It is three hops (on-ramp, on-chain, off-ramp) plus the value of time in transit. The calculator below is a StableSense teaching model: every lever is visible, and the scenario table underneath says what we are allowed to claim and from where. It is not a quote, and it is not the World Bank series.
Cost of moving $200 across borders, the World Bank Remittance Prices Worldwide measurement point. Traditional: global average ~6.49% all-in Q1 2025 Issue 53 (fee + FX margin, 1-5 days). Stablecoin gas-only: ~$0.10 network fee (seconds). The gas-only figure is the middle hop; a full fiat-to-token-to-fiat journey adds on-ramp and off-ramp FX (see the BPI 0.3-9% range in the scenario table). Race bar width is proportional to all-in cost at $200.
Cross-border cost calculator
A StableSense model of cross-border cost, not a quote and not the World Bank series. Every lever is visible. Move the sliders and watch the stacked rows match a row in the scenario table below.
Scenario table: what we are allowed to say
Each row is a named scenario. "Sourced" rows cite a primary source; "Author model" rows are clearly labeled as such. Move the calculator sliders to replay A through E.
Where adoption is real
Not pilot programs - real, non-trivial volume.
Sources: Chainalysis Geography of Cryptocurrency 2025; IMF Nigeria (Jun 2026); World Bank Remittance Prices Worldwide (Q1 2025); Bank Policy Institute (Jul 2026). Visa now settles payments in USDC (live in the US since Dec 2025).
Dollarization from below
When a country's citizens choose the dollar over the national currency at scale, monetary policy loses its transmission mechanism. Stablecoins have turned dollarization from a government-level decision into a tap on a phone. BIS Working Paper 1370 (Jul 2026) found stablecoin flows show little response to capital controls, unlike bank deposits.
Stablecoin purchases as a share of GDP. Turkey is the only country with a measured figure (~4.3%, Chainalysis). Argentina, Nigeria, Lebanon, and Venezuela have strong qualitative and survey evidence but no single comparable GDP-share figure; ranked below Turkey on documented adoption.
Three depegs, three failure modes
A depeg happens when redemption confidence drops. Start with what was supposed to hold the peg, what broke first, and whether the mechanism could recover - not with three price charts alone. Cross-linked with the dashboard Anatomy of a real depeg lesson and the in-app Research tab.
One rulebook, many flavors
By August 2026, every major financial hub has a live stablecoin framework. They converge on full reserve backing, licensed issuers, and the effective death of algorithmic designs. The real divergence is whether foreign dollar stablecoins are welcomed (Singapore) or contained to protect local currencies (EU, UK, Japan). Filter by jurisdiction below.
| Jurisdiction | Framework | Status | Permitted pegs | Algorithmic | Key rules |
|---|
Holder recovery after issuer insolvency
A reserve requirement is not the same as a holder recovery path. This table states what each jurisdiction makes enforceable, and what it leaves unresolved, as of 2026-10-18.
| Jurisdiction | Instrument and status | Reserve protection | Holder path on insolvency | Shortfall |
|---|
MiCA Article 36(2) legal segregation and the Article 39 right against reserve assets apply to asset-referenced tokens, not to the USD payment stablecoins tracked here. USDT and USDC are e-money tokens under MiCA Title IV. A reserve rule is not a recovery waterfall.
Realized case check: no fiat-reserve-backed stablecoin issuer insolvency in these four jurisdictions was found. Terraform Labs (UST) is a realized issuer insolvency with an impaired crypto-loss class, but it was algorithmic and had no reserve pool, so it does not test the reserve-pool path. The USDC/SVB event was a reserve-custodian failure, not an issuer insolvency. The absence of a case is UNRESOLVED, not proof that these paths work.
Regulatory status is the most stale-fast category in this report. MiCA enforcement dates, GENIUS rulemaking, UK Oct 2027 go-live, Hong Kong license counts all change monthly. Always timestamp.
GENIUS Act rulemaking status (as of Sep 2026)
The interest on your dollar, and who keeps it
A stablecoin issuer holds reserves that earn interest and pays the holder nothing. The gap between what the reserves earn and what the holder receives is the entire business. It is large, and the two biggest issuers report it in ways that are not comparable to each other.
Who gets paid, and whether that is new
Paying distributors for balances is not a new category of stablecoin economics. The interesting question is not whether it happens but whether anyone has compared the terms, and on that the evidence does not reach.
Tether publishes no line item called "reserve income". Its only net measure is a single line labelled "Financial result" inside the change-in-net-equity table of its attestation. Circle publishes an explicit "reserve income" figure. These are different measures, and treating them as the same number is the most common error in coverage of this market.
Tether's own documents disagree with each other on its headline. Its Q2 2026 news release leads with a positive "$1.5 billion net operating profit". Its own attestation, for the same period, reports a change in net equity of minus $3,171 million for the first half, and Q1 2026 was plus $1,040 million. That makes Q2 standalone about minus $4.21 billion. Both figures come from Tether; Tether publishes no reconciliation between them, and no Tether document defines the term "net operating profit" or states that unrealised mark-to-market is excluded. This is the single most load-bearing fact in this section, and it is invisible from either the press coverage or the release alone.
Circle, by contrast, is the simpler case and the more surprising one. Its reserve income is genuine and large, $2.64 billion for FY2025 (96.0% of its revenue), but it recorded a net loss of $69.5 million that year. Compensation expense of $844.9 million drove that, of which $424 million was stock-based compensation tied to its IPO. Almost the whole business is the float, and the pay bill took it below zero.
Machines that pay, and the three clocks they run on
This is the most-reported and least-measured part of the market. Two things are routinely conflated: whether a payment rail is live, and whether the authorization layer above it is. A rail can be settling real money while the protocol that is supposed to decide whether a machine is allowed to spend is still a draft. The table separates them.
What a headline number actually counts
Every published figure for the same activity, with its publisher and what it measures. These are not one quantity measured more or less precisely. They are different quantities, counted over different windows with different inclusion rules and different chain sets, and the spread runs from thousands of dollars to billions. Do not average them and do not subtract one from another.
The median transaction is about one cent
This is the finding the totals hide. Agent payments today are overwhelmingly tiny machine-to-machine calls, not commerce. Any comparison against card network volume is comparing a plumbing layer against retail purchases.
Who is actually paying
A payment rail can carry real money without carrying agent commerce. These are the two best attribution measurements found; each is a behavior-based bound, not a census of agent identity.
What this pass could not establish, and is therefore reported as unknown rather than estimated: any named deployment with disclosed recurring commercial scale; any first-party all-time x402 transaction count with a published methodology; any reconciliation between the headline panels, whose window, chain and inclusion rules differ (a tracker documents comparable 30-day raw totals of about $0.8M to $1.1M across dashboards); the true commerce-only share, which no source estimates on a common denominator; any production deployment of AP2; any 2026 Visa TAP or Mastercard Agent Pay transaction count; and a named publisher serving paid content over x402 in production. The x402.org "Last 30 Days" counter has shown the same four values since March 2026, so it is not a current measurement. TRM's 0.6% to 7.5% is a bound on its $25.62M screened commerce, not a census of agent identity. What the measured samples do establish is that bridge traffic and automated loops dominate: one bridging contract holds about 88% of Bitquery's August value, two loops account for 83% of the payments it found, and Chainalysis attributes growth substantially to meme-coin farming.
The chains, the consortiums, and the funds beside them
Three things sit just outside the stablecoin market and get described as though they were inside it: Circle's own blockchain, a consortium token with more than a hundred named members, and the tokenized funds that hold the same Treasuries under a different legal wrapper. Each is reported here at the level it actually exists on, which for one of them is a press release and no supply.
The funds beside the stablecoins
Tokenized funds holding the same short-dated Treasuries as the issuers, under a securities wrapper rather than a payments one. A tokenized fund share is a fund interest: the underlying securities are owned by the fund, not by the holder. Sizes are the issuers' own published figures on the date shown, because the aggregators disagree with the issuers and with each other.
BUIDL: three dated reports, no current figure
An earlier version of this report presented a $2.70-3.52B "sources disagree" range for BUIDL. A range implies a currency and a single valuation basis that none of these figures has, so they are listed separately and no current AUM is asserted.
Reported value is not market depth
Headline tokenized value counts committed and represented assets. Weekly public movement counts something else. Both are stated here, with their separate denominators, and neither is merged into the other.
How it is actually growing
The forecast chart in Section 2 points up. This is the same market measured backwards from one aggregator series, and it is the counterweight to that chart.
A rounding error, with momentum
After eleven sections of "this is a big deal," a deliberately humbling close. Today's tokenization is a rounding error against traditional finance. The growth rate is the story, not the current size.
Log scale, USD billions. US money-market funds ~$7.9T (ICI); US gold ETFs ~$530B (World Gold Council); fiat stablecoins ~$303B mid-2026 snapshot; RWA ex-stablecoins ~$20-40B (rwa.xyz); tokenized gold ~$5-8B. The ratios are structurally stable for now; the absolute sizes are fast-stale.
Until on-chain real assets approach even 1% of their traditional counterparts, claims that they are "changing" commodity or credit markets should be read as trajectory statements, not current-fact statements. The growth rates, not the sizes, are the story worth monitoring on the live dashboard.
What this edition could not close
These are open questions, not soft findings. Each needs a specific kind of evidence to close, and none is filled with an estimate. Unresolved items are stated beside the section they concern as well as listed here. What changed since the previous edition is recorded in the research changelog, not on this page.
Credit for the work that made it in
A research-contributor badge recognizes a specific contribution approved by StableSense and incorporated into published research. Each entry describes the work and links to the publication and approval record.
Have evidence, a correction or a better method? Submit it on GitHub. After review and incorporation, contributors who agree to public credit receive an entry here, a permanent link and an embeddable badge.
No contributor badges have been issued yet.
Contributors may cite and promote their specific contribution accurately. A badge does not confer project ownership, repository control, editorial authority, authorship of the whole report or endorsement. It is recognition, not a certification or a copyright transfer. Current research reuse terms remain in RESEARCH-LICENSE.
The linked contribution entry is the official record. Images can be copied or cached; check that entry for the current status.